Use cascaded pipeline to address long prediction time

[jcohenadad]

As in sct_deepseg_sc, the idea would be to first identify the ROI (eg: spinal cord, or brain), then crop the image, then only run the inference in the ROI.

Related to other issues reporting slow inference time:

• Improve the long segmentation time for spinalcord task on SCT 7.0 spinalcordtoolbox/spinalcordtoolbox#4926

Benchmarking tool:

• Benchmark inference times for DeepSeg models (CPU vs. GPU) spinalcordtoolbox/spinalcordtoolbox#4795

Recent nnunet fast cpu inference paper:

• https://openreview.net/forum?id=Zgd4DaSEW0#discussion